What it means
Ordinary databases find exact matches. A vector database finds similar meanings. After documents are turned into embeddings, the vector database stores them and, given a query's embedding, returns the closest ones in milliseconds, even across millions of items.
Options include dedicated services such as Pinecone, Qdrant, Weaviate and Chroma, and extensions that add vector search to databases such as PostgreSQL.
Why it matters for a business
Vector databases are the memory behind knowledge assistants and semantic search. For many projects, adding vector search to an existing PostgreSQL database is simpler than running a new service.
A business example
Things to watch
-
Consider your existing database before adding a new service.
-
Store source text and metadata with each vector.
-
Plan how updates and deletes stay in sync.
-
Respect permissions in search results.